and Kraft’s filtering on R2 = 0.99 but for a sibling pair plus control design. Both post-imputation filtering strategies substantially reduced the observed bias in the SNPs meeting their filter requirements, but unacceptable error rates remained (500 false positives for every 1 million imputed SNPs) (Sinnott and Kraft 2012). Moreover, correction strategies based on such highly stringent quality control metrics (i.e., R2 or RT2) require the exclusion of a large number of imputed SNPs from analyses and possibly lead to reduced statistical power and an inability to identify truly associated SNPs, especially in regions with little LD (Uh et al. 2012; Beecham et al. 2010). Applying these stringent exclusions will be particularly problematic for African-derived populations, who have shorter regions of LD across the genome.
